Open AccessAutomatic recognition of complex magnetic regions on the Sun in GONG magnetogram images and prediction of flares: Techniques for the flare warning program Flarecast
Author(s) -
Steward Graham A.,
Lobzin Vasili V.,
Wilkinson Phil J.,
Cairns Iver H.,
Robinson Peter A.
Publication year - 2011
Publication title -
space weather
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.254
H-Index - 56
ISSN - 1542-7390
DOI - 10.1029/2011sw000703
Subject(s) - magnetogram , solar flare , thresholding , flare , physics , magnetic field , sunspot , astrophysics , algorithm , artificial intelligence , mathematics , computer science , magnetic flux , quantum mechanics , image (mathematics)
In the present paper, Global Oscillation Network Group (GONG) solar magnetograms are used to automatically identify active regions by thresholding the line‐of‐sight component of the solar magnetic field. The flare potential of the regions is predicted by locating strong‐gradient polarity inversion lines (SPILs) and estimating their parameters. The parameters of interest are the length of the SPIL, a proxy for its curvature; the maximum west‐east and south‐north gradients of the magnetic field in its vicinity; and a sum of the magnetic field gradients, the summation being performed along the SPIL. Analysis for thresholding of one, two, and three parameters and the corresponding probabilities for correct prediction of flares are presented and compared. The probability for correct prediction of X‐ray flares of class C or greater in a 24 h window exceeds 88%, while the probability of false alarms is less than 10% if the decision rule involves thresholding of three specific parameters. These parameters are the steepest south‐north gradient of the magnetic field, the maximum curvature of the SPILs, and the length of the longest SPIL, all being calculated for the entire region rather than for a particular SPIL. The steepest south‐north gradient of the magnetic field is also used to estimate the probabilities for a flare to belong to classes C, M, or X. These techniques are now implemented in the flare warning program Flarecast. The first automatically predicted M‐ and X‐class flares are presented, and Flarecast is found to predict well the observed X‐ray flares.